Ministries that We Want: Parties’ Agendas and Signaling of Portfolio Salience
To measure portfolio salience over time by party, we built a panel dataset where the unit of analysis is the portfolio \(j\) as seen by party \(j\) at the time \(t\). In total, this comprises 9541 observations, 38 unique parties, and 27 unique ministerial portfolios from 2001 to 2020.
Here is a list of all portfolios and years in our dataset. As these changed over time, not only in number but in scope, we organized our list based on SIAFI’s budget data that add individual identifiers to Brazilian portfolios that make it possible to consult their spending over the years. In this way, in cases where a given portfolio changes name or structure we treat it as the same unit – in the same way Federal Deputies do regarding their decisions to allocate amendments to the budget.
To identify which parties would be included in our sample, we collected the list of all parties that had at least one seat in the Brazilian Chamber of Deputies in each year. As such, we have an unbalanced panel due to the ins and outs of parties from our sample over the years resulting from electoral losses, party switches, and other. Yet, with this operationalization we only consider parties that could produce meaningful legislative outputs signaling portfolios’ salience.
These are the party and years in our sample:
We collected data from different legislative outputs in the Brazilian Lower Chamber that, as pointed by previous studies, reveal the importance legislative party members attribute to ministries. These are:
Individual amendments to the budget proposed by Brazilian Federal Deputies – they can freely choose the ministries that end up receiving the amendments (SIAFI)
Legislative speeches by Brazilian Federal Deputies in which they mention portfolios (Câmara dos Deputados)
Legislative indications by Brazilian Federal Deputies, through which they indicate suggestions and problems to portfolios (Câmara dos Deputados)
Legislative ordinary bills by Brazilian Federal Deputies, in which they sometimes propose legislative changes to portfolios’ programs and policies (Câmara dos Deputados)
Based on this data, we defined 4 main relational indicators to measure portfolio salience.
Proportion of legislative speeches (in “Pequeno Expediente”, “Grande Expediente”, “Breves Comunicações”, or “Comunicações Parlamentares”) made by members of party \(i\) in year \(t\) mentioning portfolio \(j\) over all portfolios \(J\). To identify mentions of each one of the portfolios in our sample, we create a dictionarty of keywords containing portfolios’ names, acronyms, agencies’ names, and policies that legislators would use to refer to them. This dictionary, moreover, was organized by year – even if a given portfolio changed its name in a given year, we would still be able to detect its mentions on legislative speech. This variables can vary between 0 and 1, with mean 0.0336443
Proportion of individual amendments to the budget made by members of party \(i\) in year \(t\) to portfolio \(j\) over all portfolios \(J\). This variables can vary between 0 and 1, with mean 0.0312553
Proportion of legislative indications by members of party \(i\) in year \(t\) to portfolio \(j\) over all portfolios \(J\). This variables can vary between 0 and 1, with mean 0.0325962
Proportion of legislative bills proposed by members of party \(i\) in year \(t\) regarding portfolio \(j\) over all portfolios \(J\). This variables can vary between 0 and 1, with mean 0.0372078
The following table report descriptive statistics for our 4 main indicators of portfolio salience for parties.
In this section, we explore the relationship among our main indicators and how they vary between cabinet portfolios, parties, and years. We start with simple exercises, ploting bivariate relations and correlations among indicators before using Exploratory Factor Analysis to check if one can find a latent construct related to portfolio salience.
In the plot below, the X axis indicates the proportion of amendments to the budget submited by party to each cabinet portfolio in a given year. As the curves in the plot depicts (the dashed one represents an OLS estimate, whereas the solid one represents a LOESS), apart from the lots clustered zeros and ones, most bivariate relationships are positive – meaning that parties that submited more amendments to a certain portfolio in a given year also made more speeches about it and submited more bills and indications to it.
To inspet more formally the relationship among our indicators, here we plot the coefficients of their pairwise correlations. Reassuringly, all indicators have positive moderate association with one another, which suggest they might be useful in extracting a latent dimension of portfolio salience.
To investigate if our raw indicators of parties’ signaling of portfolio salience have face validity, we calculate the mean score of each portfolio per indicator and orderded them from top to bottom – hence, under the assumption that each indicator, measured with no error, and party has the same weight in the final mean score. Results are in plot below and show a somewhat consistent picture.
We recalculated raw portfolio salience scores by selected parties. As one can see in the next figure, these quantities vary substantively across party groups. Two facts stand out. First, there is a general agreement on the ranking of portfolios. Second, differences are consistent with parties plataform in important cases, most notably the PT with Labour and Security, or the PP with the Cities.
Below, we split these mean socoes by year to track the trend on raw portfolio salience.